]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
server: tests: passkey challenge / self-extend with context shift demo (#5832)
authorPierrick Hymbert <redacted>
Sat, 2 Mar 2024 21:00:14 +0000 (22:00 +0100)
committerGitHub <redacted>
Sat, 2 Mar 2024 21:00:14 +0000 (22:00 +0100)
commit9731134296af3a6839cd682e51d9c2109a871de5
tree882db21742d552ee948d1b5db013f02bf35ff8fa
parent4a6e2d6142ab815c964924896891e9ab3e050632
server: tests: passkey challenge /  self-extend with context shift demo (#5832)

* server: tests: add models endpoint scenario

* server: /v1/models add some metadata

* server: tests: add debug field in context before scenario

* server: tests: download model from HF, add batch size

* server: tests: add passkey test

* server: tests: add group attention params

* server: do not truncate prompt tokens if self-extend through group attention is enabled

* server: logs: do not truncate log values

* server: tests - passkey - first good working value of nga

* server: tests: fix server timeout

* server: tests: fix passkey, add doc, fix regex content matching, fix timeout

* server: tests: fix regex content matching

* server: tests: schedule slow tests on master

* server: metrics: fix when no prompt processed

* server: tests: self-extend add llama-2-7B and Mixtral-8x7B-v0.1

* server: tests: increase timeout for completion

* server: tests: keep only the PHI-2 test

* server: tests: passkey add a negative test
14 files changed:
.github/workflows/server.yml
examples/server/server.cpp
examples/server/tests/README.md
examples/server/tests/features/environment.py
examples/server/tests/features/issues.feature
examples/server/tests/features/parallel.feature
examples/server/tests/features/passkey.feature [new file with mode: 0644]
examples/server/tests/features/security.feature
examples/server/tests/features/server.feature
examples/server/tests/features/steps/steps.py
examples/server/tests/features/wrong_usages.feature
examples/server/tests/requirements.txt
examples/server/tests/tests.sh
examples/server/utils.hpp